Experimental On-Line Frequency Domain Identification and Adaptive Control of a Flexible Slewing Beam

This paper presents experimental results for the real-time adaptive identification and control of a flexible, slewing beam. A frequency domain identification algorithm incorporating non-parametric transfer function estimation and least squares parametric estimation is used to reconstruct an accurate parametric model of the system, capable of accurately tracking changing plant dynamics in real time. This model is subsequently used to produce an LQG compensator which actively damps beam vibration caused by rapid slewing manoeuvres with large payload changes. Non-persistent excitation is addressed in the context of identification during nominal motion. It is shown that after a short duration learning period, the proposed identification scheme will yield a model which is sufficiently accurate for controller synthesis.